Outcome dependent and auxiliary variable dependent sampling designs are highly efficient when compared to standard random sampling. In an era where budgets are tightening and electronic health and large-scale cohort study data are becoming increasingly available for research, study designs that use existing resources and data to advise sampling schemes are able to concentrate resources on the most informative subjects and/or times. These targeted sampling study designs (e.g., case-control and case-cohort) are ubiquitous in many areas of medical and public health research; however there has been relatively little research on designs involving longitudinal data. Longitudinal studies can address important research questions involving within- and among-individual changes in exposures and outcomes over time. The broad goal of this application is to develop a framework for conducting cost-efficient research for longitudinal data. It will include several classes of designs that are discerned by the aims, semi-parametric and likelihood based methods for analysis, and software that will permit preemptive sample size and power calculations and analyses of ascertained data. Aim 1 extends our earlier work on outcome dependent sampling (ODS) designs, to designs that sample based on multiple longitudinal outcomes. Such multiple ODS (MODS) designs enable low cost retrospective studies of pleiotropic effects of one or more expensive to ascertain exposure variables. In contrast to Aims 1, Aim 2 designs are prospective. They are conducted with outcome history and auxiliary variable dependent sampling (OHADS) schemes that alter within-subject sampling probabilities dynamically based data that have been accumulated. A primary feature of these designs is to allow researchers to weigh exposure and outcome ascertainment costs against the anticipated information gained with ascertainment at each time point. Aim 3 proposes adaptive ODS (AODS) designs that retrospectively sample subjects in waves. After each wave of subjects is collected and the data summarized, the designs are modified based on the goals of the study and will often consider estimation efficiency and robustness to modeling assumptions. In contrast to all other aims that only considered two-level data (subject and time), Aim 4 considers multi-stage outcome dependent and auxiliary variable dependent sampling (mSODS and mSADS) of hierarchical and hierarchical longitudinal data. In these designs sampling occurs are multiple levels of the hierarchical data.